AI Action Summit Learnings: Sovereign AI, regulation, and data access

Altman Solon is the largest global telecommunications, media, and technology consulting firm. In the following insight, we delve into the key takeaways from the 2025 AI Action Summit in Paris.  

The 2025 AI Action Summit in Paris, brought together stakeholders from over 100 countries to discuss a technology that is evolving at breakneck speed. Unlike previous editions, which focused primarily on AI safety, the 2025 summit covered a broader range of topics, including sovereign AI infrastructure, data availability and access, open source versus proprietary models, regulation, and AI’s growing role in crisis management.  

Sovereign AI infrastructure  

There was a palpable optimism around the potential of countries to develop "sovereign" AI infrastructure instead of relying solely on foreign AI technologies.  

The European Union is at the forefront of this movement, with a commitment to investing €200 billion in artificial intelligence, including €20 billion earmarked for AI "gigafactories." These facilities aim to advance AI research and development within the region, ensuring Europe remains a key player in the global AI landscape. 

Another significant discussion revolved around the environmental impact of AI, with France's special envoy for AI, Anne Bouverot, highlighting the growing carbon footprint of AI. France is making moves to become an emissions-free AI hub, with President Macron pledging fast-tracked data center expansion powered by nuclear energy.  

Open source versus proprietary AI models 

The debate over open source versus proprietary AI models gained significant attention. Emerging open source players from Europe and Asia are challenging the dominance of American tech giants by offering open source large language models (LLMs).  

While U.S. AI firms continue to develop "platinum-tier" proprietary models, there is increasing recognition that more accessible "gold, silver, and bronze" models from different regions can create a more balanced AI ecosystem.  These shifts could make high-performance AI tools available to a broader range of industries and organizations worldwide.  

Data: The new AI battleground 

Big Tech agrees publicly available web data is drying up. The race is on for private and exclusive datasets, driving up data costs. Access to high-quality data is becoming a critical competitive advantage.

The AI race is no longer just about computational power—it’s about data availability and access. The era of relying on vast amounts of publicly available web data is coming to an end, with tech leaders agreeing that web crawling is yielding diminishing returns. Instead, the focus is shifting towards acquiring exclusive, high-quality datasets. Proprietary data sources will be a competitive advantage. 

At the summit, multiple examples showcased how AI models trained on carefully curated datasets—ranging from 20,000 to 40,000 subjects with diverse variables—have led to significant advancements. This trend is particularly evident in healthcare, where AI is being trained on genomics, nutrition, and lifestyle data, reducing the need for excessive computational resources. 

Another major challenge in data acquisition involves data brokers and privacy regulations. Depending on geography, the legal landscape around data access varies widely, making it a complex issue for AI developers to navigate. As countries refine their data governance policies, this will remain a key area of contention in the AI space. 

Regulation: No one-size-fits-all

While global discussions on AI regulation continue, there is no universally accepted framework. Not all nations signed the "Statement on Inclusive and Sustainable Artificial Intelligence for People and the Planet," reflecting differing national priorities and regulatory philosophies. The prevailing sentiment at the summit was that a one-size-fits-all regulation is impractical, given the diverse applications and foundational models of AI. 

Instead, AI regulation is expected to evolve based on risk levels—in the same way that nuclear energy and air travel are governed under different standards. However, as of now, there is no universally recognized taxonomy of AI risks, making standardization a challenge. The development of robust testing platforms to assess AI risks will be crucial in shaping future regulations and policies. 

AI for Crisis Management

Beyond the high-level policy debates, one of the most promising trends emerging from the summit was the rise of AI-driven crisis response solutions. From mapping natural disasters in real time to humanitarian aid coordination, AI is increasingly being deployed to tackle real-world emergencies. This broadening of AI use cases signifies its growing integration into many aspects of society, from enhancing public health and safety to driving economic and environmental sustainability. 

As the AI landscape evolves, we can expect these themes to further develop and come to life over the next few years. By the time the 2026 AI Summit convenes in India, we will likely witness significant advancements and shifts in AI infrastructure, data access, regulation, and emerging applications.